The present invention relates to an article estimating apparatus that estimates articles such as daily necessities or office supplies that are handled by a user in environmental space, such as a home or an office, and also relates to an article position estimating apparatus that finds a position of the estimated article and an article estimating method, as well as an article estimating program.
Articles such as daily necessities or office supplies, located in living space, such as a home or an office (hereinafter, referred to as “environmental space”), are normally placed at various positions, such as on a desk or on a shelf. In an attempt to detect articles in such places, for example, by using an image sensor, a problem arises in which one portion or an entire portion of an object article is concealed by another article or furniture, and is not identified properly.
In order to solve such a problem, as a conventional technique for detecting an article, there has been proposed a technique of recognizing an article based upon movements of fingers trying to grab an object article (Patent Document 1). FIG. 41 is a block diagram showing the conventional article recognizing device described in Patent Document 1.
Movements of fingers to try to grab an object article are detected by a movement detection unit 1. An amount of feature is extracted from finger movement data detected by the movement detection unit 1, by a feature extracting unit 2. In a neural network unit 3, by using preliminarily set learning data, the amount of feature extracted by the feature extraction unit 2 is learned. Based upon the result of learning of the neural network unit 3, a determination unit 4 carries out a recognizing process on the object article.
Upon trying to grab an object article, a person executes an action referred to as “pre-shaping” in which a shape of a hand is prepared in accordance with the shape of the object article, and the above-mentioned method utilizes this action.